| Benchmark | Claude Fable 5 | Gemini 2.5 Flash |
|---|---|---|
| AA Agentic Index | 56.6 | 18.8 |
| AA Intelligence | 62.1 | 17.7 |
| AA-LCR | 76.7 | 65.7 |
| AA-Omniscience | 43.3 | -29.8 |
| arc_agi_1 | 90.5 | 33.3 |
| ARC-AGI-2 | 76.8 | 2.5 |
| arena_text_factuality | 1486 | 1437 |
| arena_vision | 1307 | 1214 |
| Artificial Analysis Coding Index | 76.5 | 22.2 |
| critpt | 28.6 | 1.4 |
| gdpval | 61.9 | 11.9 |
| Global-MMLU-Lite | 93.3 | 88.4 |
| GPQA Diamond | 92.6 | 82.8 |
| HLE | 64.5 | 12.1 |
| IFBench | 63.5 | 50.3 |
| MMMU-Pro | 84.2 | 69.1 |
| OmniScience Accuracy | 65.3 | 26.1 |
| OmniScience Non-Hallucination | 36.4 | 24.7 |
| scicode | 60.2 | 35.9 |
| simplebench | 81.9 | 41.2 |
| SWE-bench Verified | 95 | 60.4 |
| Terminal-Bench Hard | 62.9 | 13.6 |
| τ²-Bench Telecom (AA run) | 98.5 | 31.6 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.